Image Segmentation Tubuh Manusia Menggunakan Metode Instance Segmentation

Abstract

ImageSegmentation is theinitial step in digital imageprocessing that separatesimportant objects from the background, facilitating furtheranalysis and recognition.In human body imagery, itis crucial foridentifying and analyzing body parts such as hands, body, or head, which haveunique characteristics fordetection andrecognition. This research aims to develop a web-based system to delineate human body parts into clear segments, such as head, body, right arm, and left arm. Thestudyinvolves stages of literaturereviewto understand the concepts of instancesegmentation and YOLOv8, dataset collection, labeling, training, evaluation, andweb implementation. Thetesting results forevaluating the model's performanceshowed the highest mAPvalue of0.979 with 200epochs on thetraining dataset, 0.914 precision, 0.995 recall, and 0.95F1-score. Theconclusion of thisstudyindicates that instancesegmentation with the YOLOv8 model can provideeffectivesegmentation of human body parts.